No source can name jobs that AI will never replace, and none promises that any job is safe for good. What the evidence does show is that AI reaches tasks before it reaches whole jobs. As of 24 October 2026, the US Bureau of Labor Statistics rates hands-on, in-person work as less exposed than screen-based office work, but it warns that low exposure is not safety. Forecasts differ widely, so hold all of them loosely.
01AI changes tasks first, so 'safe' is the wrong question
A job is a bundle of tasks. AI tools can draft, summarise, sort and calculate, so they touch some tasks in many jobs and almost none in others. The International Labour Organization (ILO), a United Nations agency, estimated in May 2025 that one in four workers worldwide is in an occupation with some exposure to generative AI. It adds that most jobs are likely to be transformed rather than made redundant, because human input is still needed.
Exposure means AI could help with or do part of the work. It does not mean the job disappears. The US Bureau of Labor Statistics (BLS) says so in its AI exposure categories: exposure does not imply job loss, higher productivity, a chance of automation or a change in pay.
BLS describes its groups as relative, meaning occupations are ranked against each other. A Low label does not mean an occupation is safe from technological change.
02What the US government's AI exposure groups show
On 27 August 2026 BLS published employment projections for 2025 to 2035 and a new table that sorts 831 occupations into four groups: Low, Moderate, High and Very high exposure. The groups combine five outside studies. Three estimate in theory which tasks AI could help with. Two use real usage data from the Claude and Microsoft Copilot assistants.
Our own count of the BLS table gives 213 Low occupations and 206 in each other group. Low holds about 18 percent of 2025 jobs and Very high about 31 percent. Examples from the table:
- Low exposure: nursing assistants, janitors, maintenance workers, carpenters, construction laborers, licensed practical nurses, dental hygienists, firefighters and electrical power-line workers.
- Moderate: home health aides, electricians, plumbers, truck drivers and restaurant cooks.
- High: registered nurses, nurse practitioners, elementary and secondary school teachers, police officers and general managers.
- Very high: customer service representatives, software developers, accountants, lawyers, writers, graphic designers, data entry workers and chief executives.
BLS says the theory-based studies reflect AI as of mid-2023, and most of the studies cover language models and leave out image and video tools. The groups are a snapshot, not a verdict.
03Low exposure is not job security, and high exposure is not decline
We compared the groups with BLS's own growth projections. Of the 213 Low-exposure occupations, 78 are projected to lose jobs by 2035. Of the 206 Very high occupations, 60 are. As whole groups, Very high is projected to grow about 2 percent, Low about 2.5 percent and Moderate about 6 percent. These are our sums from the BLS table, so check them there.
The steepest falls do sit in Very high exposure. BLS projects data entry keyers down about 25 percent, telemarketers down about 21 percent and customer service representatives down about 5 percent. Yet software developers (about 10 percent growth), accountants and lawyers (about 5 percent each) are also rated Very high and are projected to grow. BLS says generative AI and automation may limit demand for office and administrative support work, which it projects to fall 4.0 percent, a loss of 752,100 jobs, the largest of any major group.
The exposure label measures what AI could do. It does not measure how many people employers will want to hire, which is why the two can point in different directions.
04What has actually happened to jobs so far
Direct evidence is thin and mixed. The Yale Budget Lab, a university policy research group, updated its tracker on 14 October 2026 and still finds no clear AI-driven shift in the mix of jobs. It also finds no link between AI usage and changes in employment or unemployment.
A Stanford team using ADP payroll data reports a different picture for the young. In a version revised on 12 August 2026, employment of workers aged 22 to 25 in AI-exposed occupations is 19 percent below where it would be had they tracked less-exposed jobs, mainly because of less hiring. The authors call the results descriptive, not causal, and experienced workers show no comparable gap.
The New York Fed reports that recent college graduates had an unemployment rate of about 5.6 percent and an underemployment rate of 42 percent in the second quarter of 2026. That page does not attribute either figure to AI.
05Why the big forecasts disagree
Headline numbers measure different things. The International Monetary Fund said in January 2024 that nearly 40 percent of jobs worldwide are exposed to AI, about 60 percent in advanced economies, and that exposure can mean either help or replacement. The ILO's 2025 figure is one in four. The World Economic Forum surveyed more than 1,000 large employers, who expect 170 million jobs created and 92 million displaced by 2030, from all structural change and not AI alone.
BLS projects total US employment to grow 3.5 percent from 2025 to 2035 and calls its uncertainty about AI very high. None of these figures is a prediction about your job. Our page on AI job loss predictions explains who made each one and how far to trust it.
06How to check any job against the sources
This is a way to read the public data yourself, not a forecast for you.
- Open the BLS table and search for the occupation by name. Note its exposure group and its projected change to 2035.
- Compare it with similar occupations in the same table, because the groups are relative rather than absolute.
- List which of the job's tasks are done at a screen and which need a person in a physical place or a person who is legally responsible. This sorting is a personal exercise, not a BLS method.
- Check again each year. BLS updates its projections annually and plans the 2026 to 2036 set for 2027.
This page is general information, not career or financial advice. For decisions about training or work, ask a qualified careers adviser.
?What is not known
- How fast employers will adopt AI at scale is unknown. BLS assumes progress follows past patterns while noting that AI is moving quickly.
- Whether weaker hiring of young workers in exposed jobs is caused by AI is disputed. Stanford calls its results descriptive, and Yale finds no clear economy-wide footprint.
- The exposure studies mostly describe language tools as of 2023 to 2025. Newer tools could change the picture within a year.
- New jobs are hard to forecast. Employer surveys count them, but task studies cannot.
QQuestions people ask
Does low AI exposure mean a job is safe?
No. In the BLS table, 78 of 213 Low-exposure occupations are projected to lose jobs by 2035 by our count. The label covers AI only, so other changes in demand are not captured.
Are healthcare jobs the safe choice?
BLS projects healthcare support jobs to grow 13.3 percent and healthcare practitioner jobs 8.0 percent from 2025 to 2035, the fastest of the major groups. Registered nurses and nurse practitioners are still rated High for AI exposure, and a projection is not a guarantee.
How often does BLS update its projections?
Every year. The 2025 to 2035 set came out on 27 August 2026, and BLS plans to publish the next set in 2027.
Last checked October 24, 2026. Rules and settings change, so open the source to see the current version.
- Employment Projections, 2025-2035 (news release) US Bureau of Labor Statistics Source for total job growth, the office support decline, the new AI exposure product and BLS's statement that AI uncertainty is very high.
- AI exposure categories US Bureau of Labor Statistics Explains how the four exposure groups were built and what exposure does and does not mean.
- AI exposure categories and 2025-35 projections (table) US Bureau of Labor Statistics The occupation-level table we used for examples and for our own counts and group totals.
- Generative AI and Jobs: 2025 update International Labour Organization Research brief used for the one in four exposure estimate and the point that most jobs are transformed.
- Tracking the Impact of AI on the Labor Market Yale Budget Lab Regularly updated analysis used for the finding that no clear AI footprint shows in the US labour market yet.
- Canaries in the Coal Mine? Six Facts about the Recent Employment Effects of AI Stanford Digital Economy Lab Working paper on payroll data used for the finding about workers aged 22 to 25 and its descriptive caveat.
- The Labor Market for Recent College Graduates Federal Reserve Bank of New York Quarterly data on graduate unemployment and underemployment, used with its second quarter 2026 figures.
- AI will transform the global economy. Let's make sure it benefits humanity International Monetary Fund IMF blog summarising its staff note, used for the exposure shares by income group.
- The Future of Jobs Report 2025 World Economic Forum Employer survey report used for the 2030 job creation and displacement expectations.
General information, not legal, medical, financial or investment advice. We are not affiliated with any company, agency or standards body named here. See our editorial policy and corrections, or tell us about a mistake at [email protected].